Eltoo

Eltoo, also written eltoo and increasingly known as LN-Symmetry, is a proposed redesign of how Lightning Network payment channels enforce their current balance. In today's channels, security rests on a penalty mechanism: if a counterparty broadcasts an outdated state to claim more than they are owed, the wronged party can punish them by taking the entire channel balance. Eltoo replaces this adversarial model with a simpler one in which any later channel state can simply override any earlier state — no punishment, no toxic old states, just "newest wins." It remains a proposal: eltoo cannot activate without a Bitcoin consensus change, and no such change has been adopted.

Symmetric state replacement

The name is a play on "L2," and the core idea is monotonically increasing state numbers. Each update transaction is constructed so it can spend the output of any prior update transaction, which means the most recent state can always be settled on-chain regardless of which stale state someone publishes first. Publishing an old state is therefore harmless rather than catastrophic: the honest party simply follows it with the newer state. Both parties hold identical (symmetric) transactions instead of the current design's per-party asymmetric pairs, which is where the LN-Symmetry name comes from.

The design was published in 2018 by Christian Decker, Rusty Russell, and Olaoluwa Osuntokun in the paper "eltoo: A Simple Layer2 Protocol for Bitcoin," and it has remained the reference answer to the penalty model's sharpest edges ever since. Settlement keeps a familiar shape: the final update pays out through a settlement transaction after a short delay window, which is what gives the honest party time to respond if a stale update ever appears on-chain.

The design has its debate. Critics note that removing penalties removes the deterrent — under symmetry, publishing an old state costs the cheater nothing but fees, so a well-resourced adversary can harass counterparties for free. Proponents answer that the attempt is also worthless: the newer state always wins, so the griefing buys nothing, and the operational safety gained — especially around backups — outweighs a deterrent that mostly punished honest users with bad disks.

Why node runners should care

The penalty model makes channel backups genuinely dangerous. Restore a Lightning node from a backup that is even one state old, broadcast what you have, and the protocol treats you as a thief — your counterparty's node will sweep the whole channel. Under eltoo, a stale backup is merely stale: the worst outcome is settling at a slightly outdated balance, not total loss. The same property lightens the load on a watchtower, which only needs to hold the latest state rather than a growing archive of penalty material for every state that ever existed. And because states are symmetric, constructions with more than two participants become far more tractable — which is why multi-party designs like the channel factory are usually sketched on top of eltoo rather than the current payment channel machinery.

What it needs from Bitcoin

Eltoo depends on a new signature-hashing mode, SIGHASH_ANYPREVOUT (BIP118), itself a proposed soft fork. Today a signature commits to the specific transaction output it spends; ANYPREVOUT lets a pre-signed transaction attach to any output whose script and amount match. That rebindability is precisely what allows one signed update to spend "whichever old state turns up." Until ANYPREVOUT or an equivalent activates, eltoo stays on the whiteboard — a well-regarded design waiting on consensus, discussed honestly as future work rather than a shipping feature. For anyone running Lightning infrastructure today, it is the clearest illustration of how one small, careful change at the base layer can remove entire categories of operational risk above it.